OpenAI Can’t Keep Up With Demand - These 4 Stocks Could Win
Milk Road Stocks
September 12, 2026
In this episode, Melvin breaks down why OpenAI’s rapidly increasing compute requirements could create major opportunities across AI infrastructure.
Speakers LGDU Set, Melvin
LGDU Set (0:00)
How much is Chat GPT actually worth? Like, is OpenAI actually a $1-2 trillion company? Many people out there are skeptical. But what if I told you that demand for their newest product, Astra, is so high that they literally have had to cut off subscriptions? Did that change your mind? And say you wanted to invest in this growth, but there's no OpenAI stock to buy, what other companies could you possibly look at? What's up, everybody? It's LGDU Set here, and welcome to Milk Road Stocks, the daily market show that likes to pretend it's not entirely about AI, but the proof is in the podcast pudding. Today is September 12th, 2026, a special weekend edition of the show recording on September 9th. OpenAI is so much more than just an LLM. It's a company with 1 billion users and a growing product in Codex that has grown from 6 million to 20 million users in just the last few months. Despite the bad press about Doomsday and a big price tag, the growth numbers are staggering. And there are a handful of satellite companies that service OpenAI that are set to benefit from this growth. Today, our top investor, Melvin, is going to take us through what those companies are and why they're so important to OpenAI's growth. And a reminder that our podcast today is free and that it wouldn't be possible without our partners at Sabre.Money, the stablecoin payments platform built for Asia. Keep an eye out for more information about them later in the show. All right, Melvin, welcome back on the show. We actually never really talk about ChatGPT and OpenAI like we do when we talk about IPOs and like some loose revenue models or revenue numbers. But I don't feel like we ever have ever spent that time diving into it. You can't buy it because they have an IPO.
If you're a freak like me, maybe you can trade some pre-shares, but that's not something anybody should do.
But I think there's a lot of things outside of it just being ChatGPT that people don't know when it comes to the money that they're actually making in the business behind it. So I'm keen to learn more, man.
Melvin (1:44)
Of course. Thanks for having me back on LG. Yeah, like LG said, today's episode is going to be a little different because we're going to be focusing on OpenAI because OpenAI is essentially back on top. They have made a huge amount of progress over the last few weeks or in a month or two.
If you look at Chat GPT, their users are growing, their Codex usage is exploding, token demand is accelerating, their advertising revenue just crossed one billion in ARR. Now, last week, Astra has officially released.
I want to give you all the stocks that are set to benefit from the growth of OpenAI. That is basically what this pod is going to be all about.
LGDU Set (2:24)
Beauty, man. Let's get into it.
Melvin (2:26)
Let's do it. Let's talk about Astra first, which is OpenAI's newest model.
Based on early benchmark results, it has taken back the lead as the best AI model in the world.
It is beating GPT 5.6, Clawed, Gemini across some of the hardest tests in math, cybersecurity, software, engineering, computer use, you name it. It's beating all of them. Because if you look at the chart on Arc AGI 3, which tests whether an AI can solve completely new problems that have never seen before, Astra scored 98.6, while GPT scored just 7.8, Claude scored 30.2, and Astra also scored 97.6 on Frontier Math, 100% on Xploy Bench, 99.2 on SRE Bench. You know, don't worry about all these means. You just have, for simplicity's sake, all you have to know is this is the best model in the world as of right now. And they did not just make small improvements. Astra is basically much better at, you know, solving unfamiliar problems like I said before, writing and fixing code, using software and completing very, very complicated tasks without, you know, really having a human in the loop. And that's what this is all about. And if you actually, let's go to the next chart. I wanna talk about the demand side of this because that'll give you a lot of context on what we're gonna talk about.
So this is the chart from OpenRouter and it shows how quickly AI usage is like increasing across the board.
OpenRouter basically tracks the number of tokens being processed across different AI models. And weekly token usage basically went from 4.6 trillion in last September to 113 trillion by August. So that's roughly 25x growth in one year. And what really stands out is that usage doubled in the last month as you can see on the chart. Now, we all know what a token is. And if you don't know what a token is, let me just give you a simple explanation. A token is basically a small piece of text or information that an AI model processes. So when people say token usage is growing, they're talking about how much work these AI models are actually doing. So for example, if I ask an AI model a short question, they only use a small amount of tokens. But if you ask it to analyze a long document, ask it to write code, fix codes, if you're using agents, those require far more tokens. And that is why this chart is important, because it shows that AI usage is not just growing, because more people are trying just one chatbot at a time, like just simple questions. But people are using these models more often, having longer conversation, giving them more complicated tasks, or using agents, as I've mentioned before. And that is the biggest increase of token usage in the last several, several months.
28 more minutes of transcript below
Thousands of transcripts fetched by people building searchable podcast archives
Fetch the whole transcript
The demo key returns a sample episode in full, no card needed:
curl -H "x-api-key: pt_demo" \
https://spoken.md/transcripts/1000651996090Markdown with the speakers named, for your notes, your knowledge base, or anything that makes HTTP calls.
From $0.10 per transcript. No subscription. Credits never expire. Prices exclude VAT, added at checkout for EU customers. Not what you expected? Email us within 14 days with 20 or fewer credits used and we refund the pack in full.
Using your own key:
curl -H "x-api-key: YOUR_KEY" \
https://spoken.md/transcripts/1000789180940